Final Report by Charlotte Burney, Ananda Costa, Breyuana Smith

Social Media Engagement & Evaluation

Table of Contents

Executive Summary◻ Questions Explored and Review of

Key Actionable Insights (3-4)◻ Summary of Research and Key

Performance Indicators (5)

Statement of Goals◻ Business Problem (6)◻ Overview of Rand Corporation (7)

2

Insights◻ Timeline Overview (9-10)◻ Owned vs Earned Data (12-19)◻ Other Social Monitoring Tools (21-22)◻ Network Analysis (24-26)◻ Content Analysis (28-32)

Recommendations◻ Future Research (34)

Executive Summary

Questions Explored in Report◻ What kinds of tweets garner the most engagement? ◻ Which subtopics of terrorism garner the most

engagement?◻ What are the most frequently discussed subtopics of

terrorism?◻ How does the conversation surrounding terrorism

differentiate across platforms?

Key Actionable Insights◻ Of the Rand Corporation's most engaging tweets, three of the five

were about terrorism, including the most engaging tweet. Meanwhile, of the ten least engaging tweets analyzed, not one of them talked about terrorism. The Rand Corporation garners more engagement when the domain links to articles about terrorism.

◻ The most engaging and least engaging tweets checked off the same boxes in terms of the content type of the tweet. They both had an image, a hyperlink, and a hashtag. Which indicates that content topic is actually more important for the RAND Corporation when creating tweets.

◻ ISIS is the subtopic of terrorism which garnered the most engagement.

◻ Facebook had the smallest amount of posts about terrorism. Twitter references mostly presidential candidates and politics in relation to terrorism. Forums discuss topics in-depth.

3

Executive Summary cont.

4

Key Actionable Insights cont.◻ The most relevant key players in this conversation are politicians

and news outlets, with Donald Trump being the most relevant influencer. The recent terror attacks have made the conversation surrounding terrorism salient in the public sphere, allowing politicians greater influence over the conversation because of the policy implications and the timing of the election year.

◻ General sentiment towards terrorism is negative except in the context of when a politician proposed how to defeat ISIS and other groups, in which case some positive feedback can be found. Surprisingly, a majority of the sentiment surrounding terrorism is neutral rather than negative, with the exception of the conversation on online forums (a 79 percent negative sentiment score). This most likely due to the informative nature of tweets about terrorism.

4

Questions Explored in Report◻ Who are the key influencers in the national terrorism

conversation? ◻ What is the general sentiment on social surrounding

the terrorism conversation?

Executive Summary cont.

Summary of ResearchTimeframe: 11/13/15 - 04/03/16

Social Listening Tool: Crimson Hexagon

Social Platforms Looked At: Twitter, Facebook, Forums

Goal: Look at the social conversation surrounding terrorism

and see how that impacts the American public sentiment

towards their sense of security within the United States.

Total Posts: 2 mil+

Other: Twitter is the only platform that had enough data to

make the gender demographic breakdown relevant

Rand Corporation tweets coded: 609

Key Performance IndicatorsMetrics: The following metrics were used to measure general

awareness and engagement of the Rand Corporation's tweets

as well as perform an extensive analysis on earned data from

the terrorism Buzz Monitor on Crimson Hexagon.

◻ Awareness: reach, impressions, volume of posts

◻ Engagement: number of active followers, likes, shares,

comments, mentions, retweets

5

Statement of Goals

Our goals for this project were to look at terrorism, specifically monitoring the current

conversation surrounding terrorism in order to measure public opinion/fears. We

understood that it would be unrealistic to try to isolate the tens of thousands of accounts ISIS

and other terrorist organizations utilize to create data sets from their accounts, therefore our

research was geared towards terrorism in the political conversation. We analyzed the owned

data from the RAND Corporation, a non-profit organization that researches and analyzes

global affairs. We looked at foreign terrorism and how it has impacted the American public

sentiment towards their sense of security within the United States. We analyzed earned data

about the social terrorism conversation across Facebook, Twitter, and Forums.

66

The Rand Corporation

7

General engagement was low on the RAND

Corporation account, with over half the tweets

generating less than 15 total engagements.

7

Understanding the Timeline8

Timeline - Crimson Hexagon

9

Initial spike in posts comes after Nov. 13, 2015 Paris attacks.

Brussels bombings on March 22, 2016.

Ankara, Turkey bombing. March 13, 2016. 37 killed 125 injured.

Obama addresses homeland security. Trump tweets about defeating ISIS. Dec. 6, 2015.

San Bernadino attacks. 14 killed and 22 injured on Dec. 2, 2015.

Timeline - Actionable Insights

10

❏ Timeline spikes correlate to events❏ Key influencers in the international

terrorism conversation cause spikes in posts

❏ Posts from Donald Trump stir and spike the conversation

Owned vs Earned Data11

Analyzing Owned Data

12

Most Engaging Least Engaging

hyperlink

image

hashtag

hyperlink

image

hashtag

All posts included: Hyperlinks, hashtags, an image, and promoted their own article

12

Analyzing Owned Data (cont)

13

Users Preferred:

Interacting with content by retweeting rather than liking because of the informative

nature of the content.

Engagement Type

Analyzing Owned Data (cont)

14

Top 3 Subtopics of Terrorism:

1. ISIS2. Domestic/Generalized

Affairs3. Paris Attacks

Engagement by Terrorism Subtopic

Analyzing Owned Data (cont)

15

Out of the 609 tweets coded, the hyperlink

breakdown showed 64% of hyperlinks tweeted led to the RAND Corporation’s

own website.

Engagement by Domain

Analyzing Earned Data

16

<1%

21%

79%

The most relevant key players in this conversation are politicians and news outlets.

Donald Trump: Most popular subtopic of the terrorism conversation.

Post Volume Breakdown

Surprisingly, a majority of the sentiment surrounding terrorism is neutral rather than negative, with the exception of the conversation on online forums.

Analyzing Earned Data (cont)

17

Trump takes up the largest part of the conversation, with approximately 25 percent of all conversation mentioning him in one way or another.

Analyzing Earned Data

18

Twitter is the only

platform with enough data to make the gender

demographic breakdown relevant.

Twitter Gender Breakdown

Important to note that all age demographic breakdowns did not have enough data to be

relevant. Neither did geographic data for Twitter.

Analyzing Earned Data

19

The subtopic of ISIS represents 59% of all

topics covered within the conversation surrounding

terrorism.Topic Wave

Other Social Media Analytics Tools20

Other Social Media Analytics Tools

21

Top Twitter InfluencersMost Retweeted Tweets

Other Analytics Tools - Actionable Insights

22

❏ Mention❏ Offers cross-platform analysis❏ Metrics include: volume over time, sentiment, geographic data for

source location, activity by day of the week, tags and top Web and Twitter influencers

❏ Not as easily navigable as Crimson Hexagon or TweetReach

❏ TweetReach❏ Would recommend it to beginners who are solely trying to grasp an

idea of what is going on with their selected topic on Twitter❏ Metrics include: estimated reach, exposure/impressions, activity, top

contributors and most retweeted❏ Not for those who are used to utilizing more complex social media

analytics software

23 Network Analysis

Network Analysis: NodeXL

Boolean Search: Israel AND terror OR tunnel

24

This Boolean search produced a large enough data sample to perform a

network analysis

User Patterns

25

Top 3 VerticesTop 3 Hubs

Network Analysis - Actionable Insights

26

❏ Due to the breaking news nature of the conversation surrounding terrorism, a lot of the users were engaged in the topic but not in conversations. This phenomenon explains the large circular community of isolates in the center of the visual from NodeXL.

❏ The largest cluster centers around Haaretz, an Israeli news source. The Haaretz had the largest in-degree centrality of all the users, most likely because it is one of the primary information sources for the population observed.

❏ Key influencers in this topic ranged from news sources, to journalists and politicians, to strongly opinionated individuals.

❏ There were no distinct large community clusters with an identifiable prefered topic of conversation. Rather, there were several medium to smaller sized clusters that shared common hashtags, such as #israel and #terror.

Content Analysis27

Content Source Breakdown

28

Twitter is the social media platform with

the most engagement.

Topic Themes by Content Source

29

Sample Posts

30

Sentiment Overview

31

Roughly 20 percent of the posts analyzed were in languages not

recognized by the sentiment algorithm used by Crimson Hexagon. These posts were

automatically filtered as neutral. It is therefore possible that the

graph is skewed neutral. However, a majority of the posts

were informative in nature, so that could also be a factor.

Content Patterns- Actionable Insights

32

◻ Forums has far different users than Twitter and Facebook and is best to be analyzed separately

◻ The conversation on Facebook about terrorism is 15 percent more negative than Twitter

◻ However a majority of the conversation surrounding Twitter and Facebook is surprisingly neutral

◻ The majority of conversation is 79% negative on Forums ◻ Donald Trump is the most popular subtopic

Moving Forward33

Resulting Actionable Future Strategies

◻ Define Boolean search after considering recent terror attacks and current top political influencers

◻ Take into consideration that while much of the social conversation occurs on forums, the users are far different than Twitter or Facebook and tailor content to each audience

◻ Redo this information in a non-election year

◻ Remember that most platforms do not have enough information to consider demographic or geographic breakdown, making it difficult to create actionable insights from the data, rather look for patterns in type of content that garners attention and engagement

34

SEESuite.uga.edu @SEESuiteUGA